Bill Inmon's Corporate Information Factory (CIF) is the top-down approach to warehouse design. You build a single, normalized, enterprise-wide data warehouse first - the integrated source of truth for the whole organization - and only then spin off dimensional data marts from it for specific departments. The warehouse is the system of record; the marts are downstream, dependent copies shaped for reporting.

Normalized 3NF at the core

The enterprise warehouse is modeled in third normal form, not as star schemas. Customer, order, and product each live in their own normalized tables with no redundant attributes. The point of 3NF here is integration: data from every source system is cleaned, reconciled, and stored once, so there is exactly one definition of "customer" and one place a value lives. This is the opposite of the Kimball approach, which models dimensionally from the start.

Dependent data marts

Departments don't query the 3NF core directly - it's not shaped for fast analytics. Instead, dependent data marts are built off the warehouse as star schemas tailored to finance, marketing, or sales. Because every mart is sourced from the same integrated core, their dimensions are conformed by construction; the conformed dimensions fall out of having one upstream source rather than being negotiated across teams. The data flow is one direction:

source systems  →  staging  →  3NF enterprise warehouse (single source of truth)
                                        │
                          ┌─────────────┼─────────────┐
                     finance mart   sales mart   marketing mart   (dependent, dimensional)

This is a hub-and-spoke architecture: the normalized warehouse is the hub, the marts are spokes.

What you gain and what it costs

The strengths are integration and consistency. With one normalized core, there is no chance two marts disagree about a metric, and the enterprise has a clean, auditable system of record - valuable in regulated industries where lineage matters. The cost is time and rigidity. You can't deliver the first useful report until a large chunk of the enterprise model is built, so the upfront investment is heavy and value arrives slowly. Schema changes ripple through the normalized core, and the 3NF structure demands strong data-modeling skill to maintain. CIF suits large organizations that need enterprise-wide consistency more than they need a quick first win.

Frequently asked questions

What is the Inmon approach? A top-down method that builds a normalized 3NF enterprise data warehouse as the single source of truth first, then derives dependent dimensional data marts from it.

Why 3NF instead of star schemas at the core? Normalization stores each fact once and reconciles all sources into one integrated model, which is what guarantees consistency. The dimensional shaping happens later, in the marts.

How does it differ from Kimball? Inmon builds the integrated enterprise warehouse first and marts second (top-down); Kimball builds dimensional marts first and integrates them via a conformed-dimension bus (bottom-up).

When is Inmon the right choice? When enterprise-wide integration, a clean auditable source of truth, and long-term consistency outweigh the need to ship the first report quickly.